DeepTutor on Render

Why deploy deeptutor on render on Render?

DeepTutor is an AI-powered learning workspace that combines tutoring, problem solving, quiz generation, research, and visualization in one system. It solves the fragmentation problem of using multiple separate tools for different learning tasks by providing an integrated, agent-native environment. This Render template deploys the full stack (FastAPI backend + Next.js frontend) as a single Docker service with secure environment-only API key management.

This template deploys DeepTutor's complete agent-native learning stack—FastAPI backend, Next.js frontend, and supervisord orchestration—as a single Docker service with a 10GB persistent disk already mounted at /app/data for knowledge bases, workspaces, and memory. Instead of wiring together the dual-process architecture, configuring the frontend-to-backend proxy, and figuring out secure secret handling yourself, everything is pre-configured with Render's sync: false secrets so your 20+ provider API keys stay encrypted in Render's env-var store and never touch disk or browser. One click gives you a working deployment with persistent storage and proper secret isolation—you just plug in the API keys you actually use.

Architecture

What you can build

After deploying, you'll have a self-hosted DeepTutor instance—a learning workspace where you can chat with LLMs, generate quizzes, solve multi-step problems, create visualizations, and run multi-agent research workflows. All your data (knowledge bases, workspaces, memory) persists on a 10 GB disk, and API keys stay in Render's encrypted environment variables, never written to disk or exposed in the UI. You need at least one LLM provider key (like OpenAI) to start; optional keys for web search, document parsing, and embeddings can be added later without redeploying the app.

Key features

• Single-container deployment: Runs FastAPI backend and Next.js frontend together under supervisord in one Docker container, simplifying deployment to a single Render web service.
• Environment-only secrets: All provider API keys are read exclusively from environment variables and never written to disk or exposed to the browser, with the render.yaml Blueprint prompting for them at deploy time.
• Persistent disk storage: A 10 GB persistent disk at /app/data stores knowledge bases, workspaces, memory, and settings that survive redeployments without including any secrets.
• Multi-provider LLM support: Supports OpenAI, Azure OpenAI, Anthropic, Cohere, and other providers with automatic detection of reasoning models that require fixed temperature settings.
• Integrated RAG capabilities: Includes document parsing, knowledge base indexing, and optional embedding/rerank services via Cohere and Jina APIs for retrieval-augmented generation workflows.

Use cases

• Instructor deploys AI tutoring workspace for students without exposing API keys
• Solo learner self-hosts quiz generation and practice tool on Render
• Developer spins up multi-provider LLM playground with one-click deployment
• Teaching assistant creates persistent knowledge base for course materials

What's included

Prerequisites

• OpenAI API Key: API key for OpenAI services including chat, embeddings, text-to-speech, and image generation.
• Anthropic API Key: API key for Anthropic's Claude language models.
• Google API Key: API key for Google AI services like Gemini models.
• Azure OpenAI API Key: API key for Azure-hosted OpenAI deployments.
• Cohere API Key: API key for Cohere embedding and reranking models.
• Jina API Key: API key for Jina AI embedding services.
• Tavily API Key: API key for Tavily web search integration.
• Brave API Key: API key for Brave Search API.
• Exa API Key: API key for Exa neural search engine.
• Serper API Key: API key for Serper Google Search API.
• Perplexity API Key: API key for Perplexity AI search and language models.
• MinerU API Token: API token for MinerU document parsing service.
• PageIndex API Key: API key for PageIndex document indexing service.
• LightRAG API Key: API key for LightRAG retrieval-augmented generation service.
• Auth Password Hash: Password hash for application authentication.
• PocketBase Admin Password: Admin password for the PocketBase authentication backend.

Next steps

1. Open the .onrender.com URL and type a question like 'Explain gradient descent' in the composer — You should see a streamed response from the default GPT model within a few seconds, confirming your OPENAI_API_KEY is working
2. Test the Solve capability by selecting 'Solve' from the composer picker and entering a math problem — You should see multi-step reasoning output that breaks down the problem and arrives at a final answer
3. Open Settings → Models and verify the OpenAI provider shows 'Set via environment (OPENAI_API_KEY) ✓' — You should see the green checkmark indicator confirming your API key is being read from the environment, not stored on disk

Resources

• DeepTutor official site